In recent years, deep learning methods have made significant progress in the field of speech enhancement, but the existing models generally have the problems of redundant parameters and high inference overhead, which are difficult to meet the requirements of edge computing devices for low resource occupation and real-time performance. In order to solve this problem, an efficient and lightweight speech enhancement framework was proposed. The front-end of the model used the band merging (BM) to compress high-frequency redundant information. Group fusion convolution (GF-Conv) was introduced, and the dilated deep convolution and time-frequency hybrid dilated residual attention module (TF-HDRA) were combined to improve the multi-scale modeling ability. At the same time, the cross-channel feature interaction was enhanced by the channel aggregation (CA) module. The grouped dual-path long short-term memory (GDPLSTM) module was used in the enhancement layer, which combined channel grouping and dual-path mechanism, which not only maintained excellent time-frequency modeling capabilities, but also realized the lightweight of the model structure, and effectively reduced the number of parameters and inference overhead. Experiments were carried out on the VoiceBank+DEMAND dataset. The results show that the voice perception quality of the proposed model is 2.99, the parameter quantity is 51 850, and the computational amount is 145.23 MMAC, which achieves a good balance between performance and computing efficiency.